Dataset Viewer
Duplicate
The dataset viewer is not available for this split.
Cannot load the dataset split (in streaming mode) to extract the first rows.
Error code:   StreamingRowsError
Exception:    TypeError
Message:      Couldn't cast array of type struct<name: string, step: int64, checkpoint: string, recipe: string> to string
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/utils.py", line 149, in get_rows_or_raise
                  return get_rows(
                      dataset=dataset,
                  ...<4 lines>...
                      column_names=column_names,
                  )
                File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
                  return func(*args, **kwargs)
                File "/src/services/worker/src/worker/utils.py", line 129, in get_rows
                  rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
                File "/src/services/worker/src/worker/utils.py", line 544, in safe_iter
                  yield from ds.decode(False) if ds.features else ds
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 3052, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2559, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2584, in _iter_arrow
                  for key, pa_table in self.ex_iterable._iter_arrow():
                                       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 558, in _iter_arrow
                  for key, pa_table in iterator:
                                       ^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 440, in _iter_arrow
                  for key, pa_table in self.generate_tables_fn(**gen_kwags):
                                       ~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 361, in _generate_tables
                  self._cast_table(pa_table, json_field_paths=json_field_paths),
                  ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
                  pa_table = table_cast(pa_table, self.info.features.arrow_schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2381, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2315, in cast_table_to_schema
                  cast_array_to_feature(
                  ~~~~~~~~~~~~~~~~~~~~~^
                      table[name] if name in table_column_names else pa.array([None] * len(table), type=schema.field(name).type),
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                      feature,
                      ^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 1861, in wrapper
                  return pa.chunked_array([func(chunk, *args, **kwargs) for chunk in array.chunks])
                                           ~~~~^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2121, in cast_array_to_feature
                  casted_array_values = _c(array.values, feature.feature)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 1863, in wrapper
                  return func(array, *args, **kwargs)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2155, in cast_array_to_feature
                  return array_cast(
                      array,
                  ...<2 lines>...
                      allow_decimal_to_str=allow_decimal_to_str,
                  )
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 1863, in wrapper
                  return func(array, *args, **kwargs)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2016, in array_cast
                  raise TypeError(f"Couldn't cast array of type {_short_str(array.type)} to {_short_str(pa_type)}")
              TypeError: Couldn't cast array of type struct<name: string, step: int64, checkpoint: string, recipe: string> to string

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

HMG Custom

326 custom G1 humanoid episodes for Box_1, Box_2 and Drill pickup, carry and placement on Table A/B. Recordings include Ego RGB, state43, action35, scene metadata, original eight-phase boundaries and skill-conditioned instructions.

Code and evaluation environments · Dataset schema · Evaluation protocols

Installation, custom-layout training and evaluation connects the official HMG environment/assets and source demonstrations to these recordings. This Hugging Face dataset is public. The GitHub code repository has separate access permissions.

Download

The current PLACE-entry training view is documented in world_he_place_entry_v1. Download the small helpers directly from this public dataset; GitHub access is not required to download the recordings or materialize portable labels:

python -m pip install huggingface_hub
python - <<'PY'
from huggingface_hub import snapshot_download
snapshot_download("xkr1706/HMG-custom", repo_type="dataset",
                  allow_patterns=["tools/download_dataset.py", "tools/release.py"],
                  local_dir="hmg-custom-downloader")
PY
python hmg-custom-downloader/tools/download_dataset.py \
  --output /data/hmg-custom --view world_he_place_entry_v1

For an existing checkout of the code repository, the historical views can also be downloaded with its helper:

git clone https://github.com/xkr1209/HMG-custom.git
cd HMG-custom
python -m pip install -e '.[data]'
hf auth login
python tools/download_dataset.py --output /data/hmg-custom --view mixed-326

No token is required for this public dataset. The full recording bundle is 43.86 GiB. The helper downloads the selected view at a pinned Hub commit, verifies the completion receipt and HDF5 checksums, and creates portable annotations/splits under training_views/<view>/. Use a new or empty output directory; interrupted downloads resume at the same revision. Original sidecars remain unchanged. Add --revision <dataset-commit-sha> for an exact release. If you already downloaded an older release, keep it intact and use a new destination for a new revision.

Dataset versions

View Episodes Full physical success PLACE-reached prefix Train / validation
table-ab-206 206 206 0 175 / 31
mixed-266 266 260 6 226 / 40
mixed-326 326 298 28 277 / 49
world_he_place_entry_v1 326 298 28 277 / 49

world_he_place_entry_v1 is a new annotation/masking view of the same 326 recordings, not another 326 episodes to concatenate. For all 28 successful-prefix episodes, it uses valid actions through PREPLACE and excludes PLACE/RELEASE beginning at the PLACE-entry boundary. This shortens 20 episodes relative to the published mixed-326 labels; the other eight partial episodes were already cut at that boundary. Raw HDF5, videos, original eight-phase boundaries and physical success/failure labels remain unchanged.

The views are cumulative, not independent collections to concatenate. mixed-326 contains 124 Box_1, 80 Box_2 and 122 Drill episodes, evenly divided between destinations within each object. Its 28 partial episodes keep physical failure labels, failed-tail masks and RGB provenance; reaching PLACE is not full physical task success.

The original table-ab-206 contains 84 Box_1 (42 per destination), 40 Box_2 (20 per destination) and 82 Drill episodes (41 per destination), totaling 164.8 minutes of interaction by control timestamps.

Observation, action and labels

Field Shape Meaning
rgb [N_obs, H, W, 3] 20 Hz Ego RGB, uint8
state [N_obs, 43] Observation-aligned robot state
action [N_obs, 35] 20 Hz observation-aligned actions
control_action [N_ctrl, 35] 50 Hz control actions; use for imitation targets
timestamp, control_timestamp [N_obs], [N_ctrl] Observation/control timestamps

Action order: waist3 + left_arm7 + left_hand7 + right_arm7 + right_hand7 + navigation3 + height. All 35 commands are retained in the recordings. Upper-body policies can supervise upper31 + height while an external Navigator owns navigation3; action35 policies can supervise navigation as well. Choose that ownership explicitly in the trainer/runtime rather than changing the recordings.

Semantic skill Source phases Example instruction
PICK NAVIGATE, PREGRASP, GRASP, LIFT Pick up the box_1.
HOLD CARRY Carry the drill to Table A.
PLACE PREPLACE, PLACE, RELEASE Place the box_2 on the table.

H30 action chunks must respect semantic boundaries and each segment's valid_end; failed PLACE/RELEASE tails are not imitation targets.

Files and integrity

Files preserve their workspace-relative data/humanoidmimicgen/datasets/ layout. metadata/views/ provides original sidecars and layout-grouped splits. Use the generated training_views/ sidecars for portable local paths. Original HMG native Box_1 supplement references are separate from these 326 custom recordings.

metadata/payload_manifest.json indexes the recording bundle. metadata/upload_content_hashes.json records every payload SHA256, including videos. upload_status.json records completion after all expected remote paths have been checked; it remains complete=false during a resumable upload.

HMG source notices are retained under licenses/HMG/; upstream robot, scene and policy resources retain their respective terms.

Downloads last month
39